Polyp-D2ATL: Deep Domain-Adaptive Transfer Learning for Colorectal Polyp Classification under Label Distribution Shift
Researchers have developed Polyp-D2ATL, a new deep domain-adaptive transfer learning framework designed to improve the accuracy of colorectal polyp classification. This framework specifically addresses challenges like imbalanced data, label distribution shifts, and cross-modality generalization. Experiments on the PICCOLO dataset showed Polyp-D2ATL outperforming existing models, achieving 82.38% accuracy and a Macro-F1 score of 77.49% on the validation set, demonstrating its clinical applicability. AI
IMPACT This new framework could lead to more accurate and reliable automated systems for early detection of colorectal polyps, potentially saving lives.